RAND Journal of Economics
Vol. 38, No. 4, Winter 2007
pp. 905–930
The rules of standard-setting organizations:
an empirical analysis
Benjamin Chiao∗
Josh Lerner∗∗
and
Jean Tirole∗∗∗
This article empirically explores standard-setting organizations’ policy choices. Consistent with
our earlier theoretical work, we find (i) a negative relationship between the extent to which an
SSO is oriented to technology sponsors and the concession level required of sponsors and (ii) a
positive correlation between the sponsor friendliness of the selected SSO and the quality of the
standard. We also develop and test two extensions of the earlier model: the presence of provisions
mandating royalty-free licensing is negatively associated with disclosure requirements, and the
relationship between concessions and user friendliness is weaker when there is only a limited
number of SSOs.
1. Introduction
The economic importance of technological standards has grown tremendously over the past
two decades. The growing recognition of the importance of the standardization process has been
attributed in large part to the growth of the information technology and communications industries,
for which standards are critical. At the same time, there has been substantial flux among these
organizations: for instance, over the past fifteen years, consortia and informal standard-setting
bodies have in many cases supplanted formal national and international standards-development
organizations (Cargill, 2002).
Because commercial stakes attached to standards and patents have become so important,
the adoption of technical approaches covered by specific patents, the requirement of backward
compatibility with earlier technologies, and the relative emphasis on cost and performance have
all been highly contentious issues. Unsurprisingly, the financial resources devoted by firms to
∗
University of Michigan.
Harvard University and the National Bureau of Economic Research; [email protected].
∗∗∗
Toulouse; [email protected].
∗
Harvard
Business School and the National Science Foundation provided financial support. We thank seminar participants
at the Federal Reserve Bank of Chicago/Kellogg “Standards and Public Policy” Conference (Chicago, May 2004),
Melbourne Business School, and the IDEI Conference on the Economics of the Internet and Software Industries (Toulouse,
January 2005), as well as Philip Haile (editor), Ken Krechmer, Mark Lemley, Halla Yang, and two anonymous referees
for helpful comments. Research support was provided by Aurora Bryant, Vicky Chang, Seung-ju Paik, Mimi Tam, and
Olga Trzebinska. All errors are our own.
∗∗
C 2007, RAND.
Copyright 905
906 / THE RAND JOURNAL OF ECONOMICS
standardization and patenting has increased sharply. 1 As discussed in more detail below, firms’
strategic behavior in standard-setting organizations has frequently been the subject of litigation
in recent years.
Despite their growing role, few statistical studies have examined differences between
the rules and operations of different standard-setting organizations (SSOs). Understanding the
workings of SSOs is important, because they can have a profound impact on the abilities of
innovators to coordinate in developing new technology, on the incentives to innovate, and so forth.
This article seeks to address this gap by investigating the relationship between these organizations’
characteristics and their policies governing the disclosure and licensing of intellectual property
such as patent awards.
Lerner and Tirole (2006) analyze forum shopping by technology sponsors. The basic idea
is that the sponsor of an attractive technology can afford to make few concessions (such as
royalty-free licensing) to prospective users and to choose an SSO that is relatively friendly to
his cause. The model thus predicts a negative relationship between the extent to which an SSO
is oriented to technology sponsors and the concession level required of sponsors, as well as a
positive association between the sponsor friendliness of the selected SSO and the quality of the
standard.
We extend this model in two ways. First, introducing disclosure policies, we show that a
higher licensing price should be associated with more disclosure. Second, we show that in settings
where there are only a limited number of SSOs, the relationship between concessions and user
friendliness may not hold: sponsor-friendly SSOs may demand substantial concessions in order to
attract weak standards; by contrast, user-friendly SSOs may make weak demands so as to appeal
to sponsors with stronger technologies. This suggests that the relationship between concessions
and user friendliness is likely to be weaker when there are fewer SSOs.
To test these predictions, we built the first database of SSOs. Combining information from
the SSOs’ websites, records of standard-setting bodies, and information collected from surveys
and interviews, we compiled a database of nearly sixty bodies.
Our results are largely consistent with theoretical predictions:
(i) First, we find a negative relationship between the SSOs’ orientation toward sponsors and the
strength of the concessions they demand. This significant negative relationship continues
to hold even when we control for industry effects.
(ii) Second, the data reveal a statistically significant association between sponsor friendliness
and the maturity of the technological subfield in which the standard is located, which we
suggest should be a proxy for attractiveness.
(iii) Third, we find that the presence of a provision mandating royalty-free licensing is negatively
associated with the presence of a disclosure requirement, whereas weaker “reasonable-andnon-discriminatory” (RND) licensing requirements are strongly associated with such a
provision.
(iv) Finally, when we divide the SSOs into those with above and below the median number
of other SSOs in their technological subfield, we find that the relationship between user
friendliness and concessions is considerably tighter among SSOs located in classes with
many other organizations.
The plan of this article is as follows. Section 2 discusses strategic interactions between firms
during standardization. The related literature is reviewed in Section 3. Section 4 presents the
theoretical framework. The data are discussed in Section 5. Section 6 presents the empirical
analysis. The final section concludes the article.
1
Jaffe and Lerner (2004) document that patent filings and litigation have increased approximately threefold in the
past two decades. IBM’s standard-development efforts are estimated to have totaled one half billion dollars in 2005 (see
www.forbes.com/2005/09/26/ibm-software-investments-cz_qh_0926ibm.html).
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 907
2. Strategic interactions and standard-setting bodies
Coordination through standardization frequently has substantial economic benefits. These include larger markets with greater economies of scale and the greater ability to sell complementary
goods. Standards can emerge in a variety of ways. First, de facto standards emerge as firms offer
competing, incompatible technologies and consumers gravitate to a particular technical solution,
such as the emergence of the Microsoft operating system. Second, some de jure standards are
selected by government agencies, as was done, for instance, by the United States in the context
of high-definition television (Farrell and Shapiro, 1992). Although government standard setting
is a fascinating topic in its own right, the very different institutional environments and incentives
suggest that it should be analyzed separately. Throughout this article, therefore, we will focus on
the de jure standardization process through SSOs organized by private parties.
The complexity of the decision-making process and the impact of standards design on firm
profitability can make the standardization process intensely competitive:
(i) Standards are frequently formed at an early stage of a technology’s evolution. In many cases,
there are a variety of promising alternatives among which the standard-setting body must
choose. These alternatives’ relative virtues may still be uncertain.
(ii) Being included or excluded from an important standard can have a substantial impact on a
firm. For instance, having one’s intellectual property deemed essential to a new standard can
help insure a steady stream of licensing revenue in future years. A standard that demands
backward compatibility can insure ongoing revenues for a legacy product for many years.
As a result, established and new firms alike are willing to devote substantial effort to the
standardization process. Standard-setting bodies have established a variety of rules to help
adjudicate this process. Industry observers often distinguish between traditional standardsdevelopment organizations, which often are open to all willing participants and have detailed,
often cumbersome procedures for adopting new standards, and special-interest groups, which
are often small, invitation-only bodies which can move rapidly to a consensus. Despite the
frequently ponderous pace at which traditional standards-development organizations move, they
are often perceived to provide a more effective “stamp of approval” than special-interest groups
dominated by technology sponsors. Looking more generally across standard-setting bodies, it
is clear that these organizations fall along a spectrum, with some bodies being more oriented
toward technology sponsors and others that reflect more the perspectives and concerns of end
users.
Because there are typically multiple technological approaches to the same problem, standardsetting groups frequently find themselves in competition, whether with other standard-setting
bodies or even with other standard-setting efforts within the same organization. To be sure,
these bodies are to a certain extent differentiated, for instance, by end-user orientation, the
standardization process followed, and the geographic composition of the membership. However,
often SSOs face the difficult choice of whether to endorse a standard developed by a different
standard-setting body, or instead to proceed with the development of a standard of their own.
A natural question is the extent to which the rules of standard-setting bodies are enforceable.
If the commitments that firms make in the standard-setting process were not binding, the value
of these commitments would be minimal. Three bodies of law are potentially relevant. 2 The
first of these is contract law. SSOs often do not require members to sign detailed contracts
stipulating their obligations relating to intellectual property. Either the firms are asked to sign
general statements agreeing to conform to the organization’s rules, or else do not need to sign
any statement whatsoever. (Formal policies were quite rare as recently as the early 1990s;
Updegrove, 2006.) Furthermore in these agreements, critical phrases—such as “reasonable”
2
This review of legal doctrine is based on Cowie and Lavelle (2002), Lemley (2002), and Mueller (2002).
C RAND 2007.
908 / THE RAND JOURNAL OF ECONOMICS
and “non-discriminatory”—are typically undefined. As a result of these considerations, contract
law is not always the mechanism chosen to enforce SSO’s rules.
Fortunately for the workings of SSOs, there are two alternatives. Two doctrines in patent
law are particularly relevant. “Equitable estoppel” covers situations where a patent holder leads
another party to believe that it will not enforce a patent against him through misleading actions.
This doctrine has been used to invalidate suits in instances when SSO members failed to disclose
relevant patent holdings. Another doctrine, termed by Lemley (2002) “implied license,” covers
situations when a firm discloses its patent holdings but then fails to comply with the restrictions
on licensing to which it concurred. Once again, the firm’s ability to extract substantial damages
from other SSO members may be limited.
As a second alternative, the members of the SSO can assert two torts against an SSO member
who behaves in an opportunistic manner. Members can claim that the failure to disclose relevant
intellectual property is a violation of the anti-monopolization provisions of the Sherman Antitrust
Act. Alternatively, they can claim that the failure to comply with the provisions of the SSO is a
form of fraud, or misrepresentation.
Most of the cases to date involving technology-based SSOs have revolved around disclosure
issues. 3 One much-discussed case involved Dell Computer’s patent covering the “VL-Bus”
standard (a bus is a mechanism to transfer instructions between a computer’s central processing
unit and its peripherals). Dell, a leading manufacturer of personal computers, was a member
of the Video Electronic Standards Association (VESA), which also included virtually all other
major U.S. computer manufacturers. In the early 1990s, the body began considering a design for
computer bus architecture that would enable faster graphics displays. VESA approved the VL-Bus
standard, which was primarily employed in the 486 family of IBM and “IBM clone” personal
computers, in 1992. As part of the approval process, representatives of member companies
certified that they knew of no intellectual property that the standard would violate.
Not until after 1.4 million personal computers were sold employing the VL-Bus standard,
the Federal Trade Commission alleged, did Dell begin contacting VESA members and indicating
that they were violating a patent that it had obtained in 1991. The FTC claimed that the standardsetting body could have readily adopted an alternative design, had they known of Dell’s patents.
Dell settled the FTC complaint in November 1995, agreeing not to enforce its patent against
manufacturers employing the VL-Bus architecture. The firm also agreed not to enforce any other
technology that it had failed to disclose to standard-setting organizations. 4
Similarly, Rambus had sued three manufacturers of dynamic random-access memory
(DRAM) semiconductors of violating its patents covering its Sync-DRAM (SDRAM) technology.
These manufacturers—Hyundai Electronics Industries, Infineon Technologies, and Micron
Technology—counterclaimed, charging the firm with having engaged in fraud and violating the
Sherman Act. In particular, they asserted that the company had failed to disclose its most relevant
patent applications to the standard-setting body in which it was an active member. Rambus
had participated in the Joint Electron Device Engineering Council (JEDEC) standard-setting
body from 1992 to one month after the Dell consent decree was finalized in mid-1996 (when
it stepped down). Rambus’s critical SDRAM patent application, which had been filed in 1990,
was not disclosed to the standard-setting body. Indeed, internal company emails referred to the
firm’s intention not to disclose this filing to the JEDEC group despite its apparent relevance.
Furthermore, executives discussed strategies to demand that other manufacturers undertake
licenses once the standard had become established. Although the initial trial found that Rambus
has indeed committed fraud, the Court of Appeals for the Federal Circuit concluded in Rambus
v. Infineon that Rambus’s obligation to disclose pending patent applications under the JEDEC
3
There has, however, been at least one recent case involving the interpretation of an SSO’s rules regarding licensing
of intellectual property (Broadcom Corp. v. Qualcomm, Inc., Civil Action no. 05-3350 (MLC), D.N.J., 2005).
4
www.ftc.gov/opa/1995/9511/dell.htm.
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 909
SSO’s policy was very narrow. 5 Because the trial record established clearly that Rambus explicitly
modified the language of its patent claims to try to cover devices with features that were revealed
to Rambus in JEDEC meetings, there can be no doubt that Rambus violated the intent and spirit
of the JEDEC policy, but not, in the eyes of the appellate court, the legal obligation to disclose.
In August 2006, the Federal Trade Commission unanimously ruled that Rambus deceptively was
able to distort the standard-setting process and “hold up” the computer memory industry. 6
3. Related literature
Despite the copious research on standards, relatively little cross-sectional work has addressed
the question of how these organizations are or should be organized. Many of the papers (e.g.,
Farrell and Saloner, 1985) focus on de facto standard setting, where there is no role for an
SSO. Alternatively, a number of works, both in economics and political science, have focused
on settings where government bodies adjudicate between the desires of different parties about
possible standards (e.g., Farrell and Shapiro, 1992). 7 In addition, several papers have considered
which settings (e.g., the extent of buyer and seller concentration and product differentiation) are
suited to the establishment of standards (for instance, Hemenway, 1975).
The economics literature on SSOs has largely focused on their role as a forum where
competitors can resolve conflicts. In Farrell and Saloner (1988), two firms can choose between two
incompatible technologies. They can do so by repeatedly talking with each other (when meeting
at the SSO), through product market competition (de facto standard setting in the marketplace),
or through a hybrid between the two approaches. The SSO in their model is a place where the two
parties can negotiate but has no institutional features (e.g., rules governing decision making or
requiring concessions from sponsors). Nor would there be a need for more than one SSO in this
setting, as the features of the SSO do not matter. The authors show that the committee process
is more likely to arrive at a high-value consensus than product market competition, but that it
usually takes longer. The hybrid approach is likely to dominate both alternatives.
Farrell (1996) models the standard-setting process as a “war of attrition” between sponsors of
two competing standards. Each sponsor, ceteris paribus, prefers her standard be selected and has
private information about its quality. Farrell shows that the higher-quality technology is ultimately
selected, but that the delay is a function of vested interests. Reducing vested interests (e.g., by
adopting rules that limit the utilization of intellectual property used in standards) reduces delay.
Bulow and Klemperer (1999) present, among other innovations, a generalized model in which
multiple firms compete to get several technologies included into a standard.
Simcoe (2007) similarly depicts the standard-setting process as a “war of attrition” between
multiple parties, each with their own proposed standard. He then corroborates the model using
standards considered by the Internet Engineering Task Force. Rysman and Simcoe (2005) stress
the importance of SSOs by showing that SSO patents are cited far more frequently than a set
of control patents, and that SSO patents receive citations for a much longer period of time.
Furthermore, they find a significant correlation between citation and the disclosure of a patent to
an SSO, which may imply a marginal impact of disclosure.
Empirical work on SSOs by economists and noneconomists alike has been dominated by
case studies. 8 Most relevant to this work is Lehr’s (1996) documentation of the intense jockeying
between firms in projects at two standard-setting bodies. Sirbu and Zwimpfer (1985) present
5
Rambus, Inc. v. Infineon Technologies AG, et al., United States Court of Appeals for the Federal Circuit, 01-1449,
-1583, -1604, -1641, 02-1174, -1192 (January 29, 2003).
6
www.ftc.gov/opa/2006/08/rambus.htm.
7
Besen and Saloner (1989) discuss nongovernmental SSOs, but they focus their more analytic discussions—
whether entailing the development of new theory or narratives that attempt to relate institutional features to theory—on
de facto standard-setting activity.
8
Lemley (2002) provides a tabulation of the policies in twenty-nine SSOs, and highlights the diversity of their
policies.
C RAND 2007.
910 / THE RAND JOURNAL OF ECONOMICS
a case study of X.25, which standardized “packet switching” over public networks. Besen and
Johnson (1986) describe seven cases where SSOs reached consensus on broadcast standards.
Weiss and Sirbu (1990) examine eleven choices between proposed standards made by standardsetting bodies. An older two-part study (1978 and 1983) by the U.S. Federal Trade Commission
also provides numerous examples. Other examples include Gandal, Gantman, and Genesove
(2007) and Rosenkopf, Metiu, and George (2001).
4. The theory
Concessions and user friendliness. As in Lerner and Tirole (2006), we explore a setting
where the owner of an idea or property must convince potential users of its value. In the bare-bones
version of the model, the utility of the users of the technology is U = a + b + c, where:
(i) a is common knowledge and measures the strength of the proposed standard.
(ii) b is unknown to both technology sponsor and users and reflects unobserved quality to users.
(iii) c is the extent of concessions made to users, such as requirements to license intellectual
property critical to the standard.
We assume that users adopt the standard only if U appears to be positive. That is, we normalize
at 0 the users’ utility either from adoption of a competing standard or from the status quo.
The attractiveness parameter a reflects, among other things, the intensity of competition with
rival technologies; a technology facing strong competition from alternative potential standards
is, ceteris paribus, characterized by a low a. The sponsor’s profit, π , is a decreasing and (weakly)
concave function of c: π (c) < 0 and π (c) ≤ 0.
The SSO chosen by the sponsor learns b and decides whether to endorse the technology. The
SSO’s objective function is a weighted average of user and sponsor benefits, U + απ . Thus, the
weighting factor, α, is (the opposite of) user friendliness. For the moment, we assume free entry
for SSOs, so there is a continuum of SSOs with different levels of user friendliness.
Let F(b) denote the cumulative distribution function. We assume that F has the standard
monotone hazard rate property: f /[1 − F] is increasing. This property in turn implies that
m(b) ≡ E[b̃ | b̃ ≥ b] grows with b at a rate lower than 1. Furthermore:
m (b) ≡
f (b)
[m (b) − b] .
1 − F (b)
Timing. We consider the following three-stage game.
(i) The sponsor chooses an SSO, that is, α, and a concession c.
(ii) The SSO learns b (more generally, it could learn a signal of b), and then chooses whether
to recommend the standard.
(iii) Users decide whether to adopt the standard.
Formally, the concession c is chosen by the sponsor. However, it can be shown that c could
alternatively be selected by the SSO; that is, under free SSO entry, there is no dissonance between
the sponsor and the selected SSO with regard to the choice of concession. 9
The SSO with type α endorses the standard if and only if
[a + b + c] + α[π (c)] ≥ 0.
The standard is therefore adopted by the users following an endorsement by the SSO if and only
if
a + E [b | b ≥ − (a + c + απ (c)) ] + c ≥ 0.
(1)
9
It can further be shown that nothing would change if c were chosen after the SSO endorses the standard and before
the users adopt the technology (see Lerner and Tirole, 2006).
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 911
The sponsor then solves
max [1 − F(−(a + c + απ (c)))]π (c)
{α,c}
subject to (1).
Proposition 1 (Lerner and Tirole, 2006). (i) The weaker the proposed standard, the more extensive
the concession and the more credible the SSO selected by the sponsor: concession c∗ decreases
and sponsor friendliness α ∗ increases with a. (ii) When concessions are (minus the level of)
royalties, that is, π = π 0 + p and c = − p, the optimal α is smaller than or equal to 1. It is equal
to 1 when the optimal royalty p is strictly positive, that is, when the attractiveness parameter
exceeds some threshold.
Intuitively, the sponsor of an attractive technology can afford both making few concessions
and choosing a friendly SSO (part i). When concessions take the form of a low royalty, but
the optimal royalty is strictly positive, the SSO is optimally balanced (α = 1) as the sponsor
captures and thus internalizes user welfare. When the optimal royalty is a corner solution at the
royalty-free level, then α ≤ 1 as the sponsor makes up for the infeasible negative royalty through
a more user-friendly SSO (part ii).
Determinants of disclosure. One aspect that was not considered in the earlier paper is
disclosure of information in the standard-setting process. In our interviews, firms highlighted
several costs associated with the disclosure of even already-issued patents. In particular, they
argued that because of the number and complexity of patent portfolios, rivals frequently could
not determine “the needle in the haystack”: that is, which patents were relevant to a given
standardization effort. 10 By highlighting the relevant patents or applications, in many cases firms
felt they were disclosing to competitors valuable information about the applicability of their patent
portfolios and their future technological strategies more generally. Second, early disclosure of
plans may invalidate the ability to get future awards. Third, the undisclosed intellectual property
may have multiple uses, only one of which is relevant to the standard. A disclosure can spur
efforts to invent around the technology and thereby either lead to a sacrifice in profits in unrelated
markets, or else boost the attractiveness of a competing standard. In such cases, the sponsor would
like to retain secrecy—or at least ambiguity—of the applicability of its patent portfolio.
Some SSOs demand that sponsors commit to revealing awards and/or applications shortly
before the standard is endorsed. Others do not require disclosure, although there is an
understanding that undisclosed patents that are later deemed relevant to the standard will be
subject to the same pricing principles as the ones that are currently examined by the SSO: for
example, a royalty-free agreement will as well cover undisclosed, essential patents in the future.
Whereas in our sample the same pricing regime (for instance, royalty-free or reasonable and
non-discriminatory) applies to both disclosed and undisclosed patents, it is not a priori obvious
why this is the case. In particular, a sponsor who does not wish to disclose patent applications,
wants to collect royalties on examined patents, and yet would like to reassure users as to the
possibility of a holdup, could offer RND on patents disclosed in advance of the adoption of
a standard, and a royalty-free treatment for undisclosed patents that are subsequently deemed
essential. 11
10
As an illustration of the disclosure effect, press releases by firms announcing patents that have already been issued
have strongly positive reactions (Erturk, Lansford, and Muscarella, 2006). In addition, U.S. legal rules mandating trebled
damages for willful infringement lead firms to discourage their engineers from even examining the patent portfolios of
their competitors.
11
Indeed, such a “mixed regime” has been proposed under the name “penalty default” by Lemley (2002). The
European Telecommunications Standards Institute had proposed in 1993 that all essential intellectual property rights of
a participant not disclosed within 180 days of the inception of a standard-setting project would be subject to automatic
licensing. This requirement was abandoned in 1995 after pressure from information technology firms (for a detailed
discussion, see Dolmans, 2002). Lichtman (2005) proposes a related mechanism called a “defensive suspension.”
C RAND 2007.
912 / THE RAND JOURNAL OF ECONOMICS
Intuitively, disclosure involves a tradeoff between reassuring users and not wasting intellectual property. On the one hand, the absence of disclosure raises the concern that users, once they
have invested in the technology, will be held up by the sponsor as a missing piece of intellectual
property is needed for the most effective implementation of the technology. On the other hand, in
the absence of holdup concerns of the users, the sponsor would prefer not to disclose applications
or technological strategies more generally.
In order to investigate the relationship between disclosure policies, pricing, and user
friendliness, let us study the following extension of the basic model. The sponsor has two pieces
of intellectual property:
(i) the existing, disclosed patent (or set of patents) that forms the basis for the standard;
(ii) an “add-on” potential patent that is subject to an application to the patent office or is merely
in the pipeline. There is no uncertainty about whether the patent on this add-on will be
granted (nothing changes if the patent will be granted with probability less than 1).
There are two states of nature, that is, two types of sponsors: “good” (probability ρ) and
“bad” (probability 1 − ρ). In either case, the add-on patent adds value H to the standard. What
differs is the baseline value. For the good type, the add-on really adds to the value of the existing
technology: the attractiveness parameter increases from a to a + H . By contrast, for the bad type,
the add-on is a missing piece in the initial standard. In its absence, the proposed standard has
attractiveness a − H only, and what the add-on does is to restore this attractiveness parameter to
its full-implementation value a. Thus, H is a measure of the potential for holdup.
The modified timing goes as follows:
(i) The sponsor applies to an SSO with parameter α and a disclosure policy specifying whether
patent applications are to be disclosed (D) or not (ND) upon acceptance. It also specifies
prices, p ≥ 0 for the basic technology and q ∈ [ 0, H ] for the add-on (whether the latter is
disclosed or not).
(ii) The SSO observes b, and under disclosure, the state of nature; it then chooses to endorse
the technology or not. If the technology is endorsed and under a disclosure agreement (D),
the sponsor must disclose the add-on.
(iii) The users choose whether to adopt the technology and, if so, pay price p.
(iv) The sponsor receives a patent for the add-on, which is then deemed essential by the SSO
(if it has not been disclosed earlier) and charges price q to the users.
The good type incurs disclosure cost d > 0 in unrelated markets, say, when the add-on is disclosed
(the disclosure cost for the bad type is irrelevant as long as it is strictly positive, as the bad type
then never has an incentive to disclose). Disclosure of the add-on reveals whether the add-on is a
true improvement or else just implementation enabling.
A couple of important points are in order. First, the sponsor cannot do better than choosing to
contract on whether to disclose and on prices (the contract is an optimal one). Because the value
added by the add-on patent is the same, H, in both states of nature, it is not possible to elicit from
users information about the state of nature. Second, we implicitly assume that the sponsor cannot
disclose to the SSO confidentially, that is, that the disclosure is subject to leakages. Otherwise,
disclosure would always be a dominant strategy for the good type (and costless for the bad type).
It would just not be perceived as costly and would be a nonissue. 12
12
The reader may wonder how the SSO can decide to endorse the standard before seeing the add-on in case the
policy is one of disclosure. This, however, is not an issue. Because only the good type may in equilibrium disclose, the
SSO can presume that the basic technology has value a and accept conditionally on checking that this is indeed correct.
Mathematically, and following the treatment below, it endorses the standard if and only if (a + H) + b − (p + q)] + α
[π 0 + (p + q)] ≥ 0, and commits not to endorse the technology if the basic value is a − H rather than the claimed level
of a.
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 913
The game is a signalling game. As will become clear from the expressions of profits, the
good and bad types have the same preferences over prices p and q. Thus, we assume that the
SSO and the users infer nothing about the state of nature from the choice of p and q. By contrast,
preferences differ as to the disclosure decision, which therefore will convey information to the
SSO and the users. Second, we will show that the signalling game always has a Pareto-dominant
equilibrium; we will accordingly focus on this equilibrium for comparative statics purposes.
Disclosure. As we noted, only the good type can benefit from disclosure. Letting b∗ denote the
SSO’s cutoff, and ( p D + q D ) the proposed prices, users adopt the technology if and only if
(a + H ) + m(b∗ ) − ( p D + q D ) ≥ 0.
Letting P D ≡ p D + q D ≥ 0, the sponsor’s expected profit is
[1 − F (b∗ )] π0 + P D − d ,
accounting for the fact that the disclosure cost d is incurred only in case of endorsement, as is
the case in practice. (Note, too, that such conditional acceptance maximizes the SSO’s appeal, as
it avoids wasteful disclosure when the standard is turned down, that is, when b < b∗ .)
The good type’s profit under disclosure is then
π D (d, a) =
max
{b∗ |b∗ ≥m −1 (−(a+H ))}
{[1 − F (b∗ )] [π0 + (a + H ) + m (b∗ ) − d]} ,
where the set restriction refers to the constraint that the price P D be nonnegative. 13
Nondisclosure. Let ρ̂ (0 ≤ ρ̂ ≤ ρ) denote the SSO’s and the users’ posterior probability of the
sponsor being of the good type when there is no disclosure. Letting PND ≡ pND + qND , and b̂
denoting the SSO’s cutoff, users adopt the standard if and only if
(a + ρ̂ H ) + m(b̂) − P N D ≥ 0.
The sponsor’s expected profit in the absence of disclosure is type independent and equal to
π N D (ρ̂, a) =
max
{b̂|b̂≥m −1 (−(a+ρ̂ H ))}
{[1 − F(b̂)] [π0 + (a + ρ̂ H ) + m(b̂)]}.
A separating equilibrium exists if and only if
π D (d, a) ≥ π N D (0, a) .
A pooling equilibrium exists if and only if
π N D (ρ, a) ≥ π D (d, a) .
Finally, a semi separating equilibrium exists if there exists ρ̂ ∈ (0, ρ) such that
π D (d, a) = π N D (ρ̂, a) .
Lemma 1.
(i) If multiple equilibria coexist, both types of sponsor are better off in the one with the
least amount of disclosure (the maximal amount of pooling). Furthermore, either the
separating equilibrium exists and is unique, or the Pareto-dominant equilibrium is the
pooling equilibrium.
(ii) Focusing on the Pareto-dominant equilibrium in case of multiplicity) There exists
d ∗ > 0 such that separation obtains for d < d ∗ and pooling for d ≥ d ∗ .
13
If at the optimum P D > 0, then there is an indeterminacy as to the respective levels of p D and q D (as long as
p D + q D = P D and q D ≤ H ). If there were a cost of developing the add-on, arbitrarily small in expectation but with
wide support, the optimal contract would backload payments through a two-part tariff with q D = min {P D , H }, so as to
provide the sponsor with maximal incentives to develop the add-on. The indeterminacy would be removed.
C RAND 2007.
914 / THE RAND JOURNAL OF ECONOMICS
Proof .
(i) π D is belief free. By contrast, the two types’ (common) payoff in the absence of
disclosure is increasing in ρ̂. Therefore both prefer ρ̂ to be equal to ρ. More formally,
either π D (d, a) > π ND (ρ, a) and then the equilibrium is unique and separating, or
π D (d, a) ≤ π ND (ρ, a) and then the pooling equilibrium exists and dominates any other
equilibrium.
(ii) This results from the fact that π D is decreasing in d, whereas π ND does not depend on
d.
As already mentioned, we will select the Pareto-dominant pooling equilibrium when it exists
(when it does not, the separating equilibrium is the only equilibrium anyway).
Proposition 2 (within equilibrium: disclosure positively correlated with (α, P)). In (a separating)
equilibrium, disclosure is associated with (i) higher prices and (ii) lower user friendliness of the
selected SSO.
Note that we focus on separating equilibria. There is no variation if pooling obtains.
Intuitively, disclosure demonstrates the absence of holdup and thereby makes the technology
more attractive. Proposition 2 therefore is akin to Proposition 1 (the proofs of Propositions 2–4
can be found in the Appendix).
The next proposition shows that disclosure is akin to a concession in that more attractive
standards are less likely to require disclosure.
Proposition 3 (disclosure is less likely for an attractive standard). When the technology becomes
less attractive (a falls), the range of disclosure costs for which disclosure occurs in equilibrium
expands (d ∗ grows).
Positioning with limited SSO competition. The analysis so far has made the extreme
assumption that there is free entry into the SSO market and delivered a number of sharp
implications, including the negative relationship between sponsor friendliness and concessions.
A finding that α and c covary negatively, though, might be attributed to the possibility
that there is a limited number of SSOs for a given technology field and that user-friendly SSOs
just demand more concessions. To assess the validity of this alternative theory, it is important
to distinguish two types of SSO policies: ex ante rules and ex post discretionary actions. Userfriendly SSOs will naturally ex post demand more concessions. 14 On the other hand, our empirical
analysis focuses on the ex ante rules that govern applications to the SSO. As we now show, it is
much less obvious that a more user-friendly SSO with market power will choose tougher rules.
To illustrate this, we look at the case in which a fixed set of SSOs selects concessions in
order to attract a sponsor with known characteristic a (and therefore preferred SSO α ∗ (a)). For
expositional simplicity, we ignore disclosure decisions and return to the basic framework at the
beginning of of Section 4.
Our notion of competition can be interpreted as one among either for-profit SSOs that
maximize revenue or not-for-profit SSOs that try to attract business; in either case, the SSO
chooses c so as to solve
(I )
max U (α) = [1 − F(b)]π (c)
subject to
a + m(b) + c ≥ 0
14
Whether they will indeed be successful in their attempt at “technology morphing” is another matter. To the extent
that they delay approval, they may signal bad news to the users and so compromise the very acceptance of the standard
by users even with increased concessions; see Lerner and Tirole (2006).
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 915
and
a + b + c + απ (c) = 0.
(A not-for-profit certifier tries to attract the sponsor’s business; a for-profit one charges P(α) =
max{U (α) − maxα =α U (α ), 0}; in either case, the certifier wants to maximize U (α).)
Proposition 4 shows that over a range of parameters, the concession is then increasing with
sponsor friendliness. It therefore precludes any general conclusion as to the negative covariation
between α and c. Sponsor-friendly SSOs are ex post lenient (with regard to their choice of b∗ ) and
so must impose strong concessions (high c) in order to persuade consumers to adopt technologies
that have weak appeal, and thereby attract such technologies. User-friendly SSOs must demand
low concessions in order to be attractive to sponsors of technologies with strong appeal.
Proposition 4 (SSO positioning under imperfect SSO competition). Consider the competition
between a given set of SSOs for a sponsor whose technology has appeal parameter a and preferred
SSO is thus α ∗ ≡ α ∗ (a). There exists α < α ∗ such that for α ≥ α, the concession c made by SSO
α is increasing with α. That is, the sponsor must make fewer concessions when applying to the
more user-friendly SSO.
5. The sample
Before testing the theory, we describe the construction of our sample. Our empirical approach
was to develop as comprehensive a sample of SSOs as possible. To do this, we compiled data
ourselves and also searched the Internet for information about these organizations and their
policies.
To identify the SSOs we would employ in our study, we employed “snowball sampling,” a
popular form of non-probability sampling. Krathwohl (1997) notes that “snowball sampling is
used to discover the members of a group of individuals not otherwise easily identified by starting
with someone in the know and asking for referrals to other knowledgeable individuals.” We first
started off with the list in Lemley (2002). Lemley limited his population to groups that (i) disclosed
their intellectual property policies in online documents or after direct contacts and (ii) included
Sun Microsystems as a member or an observer in the late 1990s. His judgment is that this method
is representative of industries that he terms “telecommunications and computer networking
industries,” in which the most contentious IP issues arise. From this paper, we found a reference to
another list at consortiuminfo.org. We employed all SSOs listed at consortiuminfo.org that related
to information technology and communications technologies and disclosed their intellectual
property policies in online documents or after direct contacts, including those that Lemley did
not use (apparently because his second criterion was not satisfied). We also used a variety of other
lists, including www.cenorm.be/isss/Consortiua/Surveyshort.htm, www.diffuse.org/fora.html,
www.webstart.com/cc.standards.html, 15 and www.marinade.ltd.uk/content/standard.html. These
additional lists had diminishing returns, and we stopped when we could not identify any additional
lists which had any standards that met our criteria.
We then coded the characteristics of the organizations based on the data collected from
publicly available sources such as bylaws, charters, and websites of these bodies, as well as the
authors’ survey, emails, and telephone interviews. 16 In some cases, observations needed to be
dropped after we could not obtain consistent answers to our questions.
Our sample encompassed technologies related to the information technology, telecommunications, and electronics industries. More specifically, in the ISO Catalogue, each ISO standard
is listed under a technological subfield such as “computer graphics,” which is in turn under a
technological field such as “information technology and office machines.” All but one of the SSOs
15
The link is now changed to www.cmpcmm.com/standards.html.
We created a survey website in June 2003. It is available at cess.nyu.edu/hfc/sso. We thank ISO for allowing us
to base our survey on the ISO classification. The databases we used included those of Gale and ILI Infobase.
16
C RAND 2007.
916 / THE RAND JOURNAL OF ECONOMICS
in our sample fell into three technological fields (field number in parentheses): (31) electronics;
(33) telecommunications, audio, and video engineering; and (35) information technology and
office machines. In fact, 80% of the sample (47 out of 59) fall exclusively in these fields.
Although non-probability sampling is very common because of its convenience, it poses a
number of concerns. First, there is no clear-cut way to tell how representative in the statistical
sense the sample is. Because there are relatively very few empirical works on SSOs and there
is no comprehensive list containing all such organizations, however, we could not find a better
alternative.
To answer the question about whether our sample is big or small relative to the total, we
provide here some references for comparison:
(i) www.cmpcmm.com/standards.html lists 141 entries of both standards and SSOs, a significant portion of which are standards, not SSOs.
(ii) consortiuminfo.org lists 26 national, 10 international, and 11 regional SSOs.
(iii) ILI standards database, which covers over 600,000 worldwide standards across many
different technologies, claims to include over 250 major standards-issuing authorities, from
the United States, Europe, Japan, Australia, and major international bodies. 17
(iv) PERINORM covers over 500,000 worldwide standards. It includes 21 national institutes. 18
A second concern is our inability to observe SSOs that had failed prior to the data-collection
effort. All the SSOs in our sample were still in existence at the time of the data-collection effort,
because those are the organizations that we could still collect data on their policies. (Failed SSOs
typically do not maintain a website that allows one to learn their policies or key contacts.) A
problem with such bias is that there could possibly be some characteristics of those organizations
that could explain why they failed or had a short life which are not captured in our sample.
Related to the sample bias question is the question of the independence of data. If, for
instance, U.S. SSOs overwhelmingly followed the template developed by the American National
Standards Institute—the body that, among other activities, accredits standards developers and
represents the United States in the International Organization for Standardization—the statistical
significance of our findings might be overstated. In fact, intellectual property policies of SSOs
are very diverse. For instance, Lemley (2002) writes: “What is most striking about the data is the
significant variation in policies among the different SSOs.” This claim is illustrated in Table 1:
along the dimensions summarized here (a subset of the intellectual property policies of SSOs),
there are only two organizations (namely, DVB and IrDA) that have the same policies as ANSI.
Table 1 provides an initial overview of the sample. It lists the names of the organizations,
their websites, and some additional information, such as:
(i) Whether the organization has a policy covering patents.
(ii) The rules regarding the licensing of patents by the SSO. In particular, we highlight whether
the SSO requires sponsors to commit to license this intellectual property on RND terms, on
a royalty-free basis, and two less-common variants: provisions that the sponsors assign their
intellectual property to the SSO and that the SSO can compel licensing of the sponsors’
patents.
(iii) Whether sponsors must commit to abide by a formal dispute-resolution process.
(iv) Whether there are requirements to disclose relevant patents (and in some cases, applications)
before the selection of the standard.
It should be acknowledged that the uncertainty and pending litigation alluded to above
implies that our characterization of SSOs’ licensing rules is likely to face an errors-in-variables
17
18
See englibrary.blogspot.com/2006/03/ili-standards-infobase.html.
See www.cssinfo.com/perinorm.html and www.bsi-global.com/Business_Information/Publications/perinorm.
xalter.
C RAND 2007.
Address
www.ansi.org
www.atmforum.com
www.bcdforum.org
www.bioapi.org
www.bpmi.org
www.bsi-global.com
www.cenorm.be
www.cpexchange.org
dublincore.org
www.dmtf.org
www.dvb.org
www.ecma.ch
www.ectf.org
www.edifice.org
www.etis.org
www.etsi.org
www.frforum.com
www.gigabit-ethernet.org
www.homeplug.org
www.homepna.org
www.homerf.org
www.i2osig.org
www.ieee.org
www.ietf.org
www.imtc.org
www.internethomealliance.com
www.irda.org
www.iso.ch
www.itu.int/ITU-T
ANSI
ATM Forum
BCDF
BioAPI
BPMI
BSI
CEN
CPExchange
DCMI
DMTF
DVB
ECMA
ECTF
EDIFICE
ETIS
ETSI
Frame Relay Forum
GEA
Home Plug
Home PNA
HomeRF
I2O
IEEE
IETF
IMTC
Internet Home Alliance
IrDA
ISO
ITU-T
Sample Overview
Standard Development
Organization
TABLE 1
C RAND 2007.
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
NA
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Patents Covered by
Organization Policy
Y
N
Y
Y
N
N
Y
N
N
Y
Y
Y
Y
NA
Y
M
Y
Y
Y
Y
Y
N
Y
Y
N
N
Y
Y
Y
RND
N
N
N
Y
N
N
N
N
N
N
N
N
N
NA
N
N
N
N
N
N
N
Y
N
N
N
N
N
N
N
Royalty-Free
N
N
N
N
N
N
N
Y
N
N
N
N
N
NA
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
Assignment
N
N
N
N
N
Y
N
N
N
N
N
N
N
NA
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
Compulsory
External Patent-Licensing Rules
Y
N
N
N
N
Y
N
N
N
N
Y
N
N
NA
N
Y
N
N
M
N
N
N
N
N
N
N
Y
N
N
Dispute Resolution
Mechanism
Continued.
Y
Y
Y
Y
N
N
Y
N
N
Y
Y
Y
Y
NA
Y
Y
Y
Y
N
Y
N
N
Y
Y
Y
N
Y
Y
Y
Disclosure
Requirements
CHIAO, LERNER, AND TIROLE / 917
Continued
C RAND 2007.
MEF
MSF
MWIF
NMF (Telemanagement Forum)
NPF
NSIF
OASIS
OGC
OIF
OMA
OMG
Open Group
OSDL
OSGi
PCCA
PCI SIG
RDMA
RosettaNet
SDMI
SNIA
STA
TIA
UPNP
W3C
WAP Forum
WfMC
Wired for Management
X.org
XIWT
XML.org
Standard Development
Organization
TABLE 1
www.metroethernetforum.org
www.msforum.org
www.mwif.org
www.nmf.org
www.npforum.org
www.atis.org/atis/sif/sifhom.htm
www.oasis-open.org
www.opengis.org
www.oiforum.com
www.openmobilealliance.org
www.omg.org
www.opengroup.org
www.osdl.org
www.osgi.org
www.pcca.org
www.pcisig.com
www.rdmaconsortium.org
www.rosettanet.org
www.sdmi.org
www.snia.org
www.scsita.org
www.tiaonline.org/standards
www.upnp.org
www.w3.org
www.wapforum.org
www.wfmc.org
developer.intel.com/ial/wfm/wfmspecs.htm
www.x.org
www.xiwt.org
www.oasis-open.org
Address
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
NA
Y
Y
Patents Covered by
Organization Policy
Y
Y
N
Y
Y
Y
N
Y
Y
Y
Y
Y
N
Y
N
Y
Y
N
N
Y
N
Y
Y
N
Y
N
N
NA
N
N
RND
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
Y
N
N
N
N
N
Y
N
Y
Y
NA
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
NA
N
N
Assignment
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
NA
N
N
Compulsory
External Patent-Licensing Rules
Royalty-Free
N
N
N
N
N
N
N
M
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
NA
N
N
Dispute Resolution
Mechanism
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
N
NA
Y
N
Disclosure
Requirements
918 / THE RAND JOURNAL OF ECONOMICS
CHIAO, LERNER, AND TIROLE / 919
problem. Although the formal licensing policies can be observed, the extent to which these rules
are enforceable remains in question. The Rambus decision discussed above and several others
have engendered doubts about the enforceability of SSOs’ rules about intellectual property.
6. The analysis
We next present the analysis. First, we discuss the proxies that we have developed for c, the
measure of concessions demanded by the SSO, and α, the orientation of the SSO to sponsors
relative to users. We then explore their relationship. We also consider the relationships between
α, disclosure requirements, and the maturity of the technology.
The relationship between user friendliness and concessions. Table 2 summarizes the
elements of the two indexes. The first seeks to capture α, the extent to which the SSO is oriented
to users or sponsors:
The nature of the organization. Special-interest groups (SIGs) are frequently observed to have
a greater orientation to sponsors than other organizations. The membership of these groups is
frequently confined to intellectual property rights holders. 19 Moreover, SIGs’ mandates frequently
also include the marketing of these standards rather than just a dispassionate endorsement. The
contrast is sharpest between SIGs and standards-development organizations (SDOs) such as the
Institute of Electrical and Electronics Engineers. An SIG has a narrower interest than an SDO.
Also, the specifications of an SIG are more likely to come from a single member, whereas SDOs
build standards based on the contributions of various members. 20
The nature of the membership. Some SSOs have individual members, whereas others are
confined to corporations. (Yet others involve additional parties, such as academic institutions
and government agencies. 21 ) We regard SSOs with all-corporate membership as higher-α
organizations. 22
The nature of the voting rules.23 In interviews with a number of practitioners, the importance of
voting rules was highlighted. The SSOs that required standards to be approved by consensus or
with a super-majority were seen as being much less prone to endorse a standard than those that
use majority voting. 24
19
In an illustrative email, one informant writes that “the working groups of our organization are comprised of
members from many of the large industry leaders—therefore many of the companies that have an interest in IP protection
are already stakeholders.”
20
It should be noted that there is also a third class of organization. Fora, inasmuch as the organizations are involved
in the standard-setting process, are frequently seen as a middle ground. As a platform for the exchange of information,
members of fora facilitate, accelerate, and promote the general interoperability of products in an industry. Fora work
with other SSOs to develop standards and improve the usability of standards by preparing implementation guidelines as
recommendations to members on the usage of a standard. Although these organizations often seek to make use of existing
standards whenever possible, they may also create their own standards or specifications. Fora thus can be seen as both
complements to and substitutes for other SSOs.
21
More than half of the organizations (57%) consist of corporations only. Eight percent consist of both individuals
and corporations, and 25% consist of corporations and others. One organization consists of all three types of members.
Almost all organizations (92%) have corporate members.
22
Our results continue to hold when we include organizations whose members are exclusively governmental bodies
(which might be prone to pressure from national corporations) with the all-corporate firms.
23
The major sources of information here are the charters and bylaws on the websites of the organizations, and the
survey. Sometimes, there is no specific decision process pertaining to standard setting. In that case, we assumed that the
publicly available decision process, which pertains to general decisions, encompasses standard-setting decisions also. If
the organizations answered the voting rule question in our survey, the answers were used to compile the data. Otherwise,
we compiled the data ourselves by reading the charters and bylaws of these organizations; we created a summary for each
of these organizations. The summaries are archived at cess.nyu.edu/hfc/sso/decisionprocess.zip.
24
Of organizations, 34% and 27% use majority voting rules and super-majority to approve standards, respectively;
13% of organizations use consensus. There is no information for the remaining 25% of organizations.
C RAND 2007.
C RAND 2007.
∗∗∗
15
80
26
0
60
20
denotes significant at the 1% confidence level; ∗ ∗ , 5%; ∗ , 10%.
14
83∗∗∗
No (%)
10
40
Yes (%)
SIG
(SIGs and SSOs Only)?
60
20
0
10
40
Yes (%)
81
24
11
13
80∗∗∗
No (%)
SIG (All)?
75
26
8
0
59
Yes (%)
79
21
10
24∗∗∗
87∗∗
No (%)
All Members
Corporate?
79
32
0
5
65
Yes (%)
76
19
14
16
77
No (%)
Decisions Made
by Majority Rule?
α Score Element (Higher α Choice Is the Left Column)
χ 2 and t-Test Analyses of Relationship between α and c Score Elements (with Disclosure Measures)
c Score elements
Royalty-free licensing?
Royalty-free or
RND licensing?
Binding dispute resolution?
Disclosure measures
Is patent disclosure required?
Is application disclosure required?
TABLE 2
68
8
0
7
69
Yes (%)
0.6
1.3
0.4
1.4
1.5
18∗∗
86∗
30
Yes (%)
1.9
1.5
1.6∗ ∗
1.6∗ ∗
2.1∗∗∗
No (%)
α Score (0–4)
If Element Is. . .
17
77
No (%)
Younger
Organization?
920 / THE RAND JOURNAL OF ECONOMICS
CHIAO, LERNER, AND TIROLE / 921
The age of the organization. Numerous observers (e.g., Cargill and Bolin, 2007) have observed
that the standard-setting process has become increasingly politicized over time. These observers
have attributed this trend to the growing involvement of lawyers and business-development
personnel in an activity that had been previously dominated by engineers. (See Simcoe, 2007
for empirical support of this claim.) Although many established organizations have adjusted
their rules to accommodate the interests of the sponsors, these changes have frequently been
slow. Discussions suggest that rules have been slanted in a more pro-sponsor direction most
dramatically in the more recently established SSOs. (These newer organizations have lacked the
institutional traditions that have served to slow the pace of change in older groups. 25 ) We expect
SSOs established after the median date in the sample (1995) to have a higher α.
It is difficult to assign a relative importance to these four elements. We thus simply—in an
admittedly imperfect approximation—sum these four dummies, and create an α score between
zero and four. In diagnostic regressions below, we also look at each element separately.
We similarly create an index of c, the number of concessions offered by the sponsor. We focus
on the two elements identified as most critical in our discussions, the commitments regarding
licensing and the allocation of residual decision-making rights: 26
Licensing restrictions.27 In a number of SSOs, firms must commit to license key intellectual
property needed to implement the standard to those who request it. These commitments typically
take two forms: the firm commits to license the patents either on a royalty-free basis or on RND
terms. Although ambiguities surround exactly how binding a RND licensing commitment is, most
observers see it as a serious commitment, though clearly not as restrictive as a promise to provide
royalty-free license (see Lemley, 2002). We include two dummies: one that takes on the value
one if the firm must commit to provide royalty-free licenses, and one that takes on the value one
if they must commit to either RND or royalty-free licenses. 28 We include compulsory licensing
and patent assignment requirements as equivalent to royalty-free licensing for the purposes of
this analysis. 29
Residual decision rights. In many SSOs, there is no clear road map to resolving disputes. In
others, however, the firms must commit to bringing their disputes before an adjudicary body of
the SSO. 30
Again, we create an index of concessions, which ranges from zero to two.
25
All of the organizations adopting the consensus rule are significantly older than the majority of organizations.
Note that nearly half of the organizations in our sample were founded between 1996 and 2002, and 73% of them are less
than fifteen years old.
26
Organizations vary in the extent to which they disclose historical information. Although some organizations
carefully archive earlier policy documents, in many cases only contemporaneous policies are available. Thus, when
coding these policies, we simply focused on the policies that were in place in October 2002. We ignored proposed
alterations to these policies when available in draft form, focusing instead on policies actually in place.
27
Organizations typically define (i) the rules (e.g., RND licensing requirements) governing the intellectual property,
and (ii) the range of covered intellectual property. Almost all organizations (96%) include patents among the intellectual
property policies covered under their rules. About half of the organizations (45%) have policies governing trademarks,
77% copyrights, and 39% other types of intellectual property rights.
28
It is worth highlighting that many crucial details are often not stipulated in these contracts. The case Intel v. VIA
Technologies (174 F.Supp. 2d 1038 [N.D. Cal. 2001]), for instance, revolved around the question of whether the licensing
commitment entered into as part of a standard covered just the basic features of the standard or else also included various
extensions. Some of the loopholes that firms have successfully exploited are cataloged in Feldman, Rees, and Townshend
(2000) and Kipnis (2000).
29
The majority (63%) of organizations use RND in their patent-licensing rules. Only 9% of organizations use
royalty-free rules. Even fewer organizations use assignment (2%) and compulsory rules (2%). We also repeat the analysis
with a third dummy that takes on the value one for those SSOs that require patent assignment (which might be seen as
particularly taxing). The results are little changed.
30
Our data show that only 9% of organizations have a dispute resolution mechanism. It should be noted that in
some cases, separate provisions govern copyright licensing, but we have not recorded these.
C RAND 2007.
922 / THE RAND JOURNAL OF ECONOMICS
The final two measures reported in Table 2 relate to the extent of disclosure in the SSO. As
highlighted in the second subsection of Section 4, the predictions here will be somewhat different
from those for other concessions because of the adverse selection effect. We report whether the
sponsors were required to disclose either patent awards or applications prior to the adoption of a
standard.
Table 2 reports the results of this analysis. Although not all cross-tabulations between the
proxies for α and c are statistically significant, the basic pattern is clear: in every case, the level
of concessions, c, is higher for those with a lower proxy for α, that is, a greater orientation toward
users. When we add together the α dummies in the final two columns, we see in each case the
difference is significant at the 5% confidence level. For instance, for SSOs that require royalty-free
licenses, the α score is 0.6; for the others, it is 1.6. For SSOs requiring binding dispute resolution,
the α score is 0.4; for the others, it is again 1.6.
The results regarding the disclosure requirements are much less clear-cut. Only one of the
ten cross-tabulations is significant, and that only at the 10% confidence level. The summed α
scores are also not significantly different from each other.
In Table 3, we look at the correlations between the α and c proxies. There is a strong
negative correlation between the scores, −0.53, which is highly statistically significant: that is,
more user-friendly SSOs offer more concessions. We also examine the correlation between these
two scores and the individual elements of the other index. These correlations are each statistically
significant, at least at the 10% confidence level. Once again, no significant relationship between
the disclosure policies and the α score appears.
We next turn to regression analyses. In Table 4, we seek to explain the extent of concessions
offered by an SSO, given its level of user friendliness. (This assumption of the exogeneity
of α is plausible if we assume free entry. As the tabulations of the frequency of other SSOs
in the technological subfields of our sample discussed below reveal, this assumption does not
appear unreasonable. If it did not hold, we should probably regard this as more representative
of correlation than causation.) In each case, we estimate first a basic specification, and then
with (unreported) dummy variables that control for the technologies covered by the SSO. (We
determine these first by checking the ILI Infobase, which classifies the standards published by the
SSOs according to the ISO technological fields—a scheme used in the International Organization
for Standardization’s ISO Catalogue. 31 For SSOs not included in the database, we asked the
organizations to respond to our survey, filling out the number of standards they published in each
field and subfield. For nonresponding SSOs, we make our own classification, based on information
from the mission statements or elsewhere on the organizations’ websites.) We employ ordered
logit regressions throughout, reflecting the fact that although we expect that an organization with
a c score of 2 is more restrictive than that with a score of 1, it is difficult to say exactly how much
more restrictive it is.
The table reveals again that there is a strong relationship between c and α, even after we add
industry controls. When we examine the individual elements of α, we see that whereas all the
coefficients are negative, the two consistently significant indicators are if the standards body is
an SIG and if all the members are corporations.
When we look at the determinants of the three individual components of the c score in
Table 5, we see that the α score is consistently negative in each. Moreover, each coefficient is
statistically significant at least at the 5% confidence level. When we look at the two disclosure
measures, we find that not only is the α coefficient statistically insignificant but that it takes on a
different sign.
31
This is available at www.iso.org/iso/en/CatalogueListPage.CatalogueList. See also “How Are ISO Standards
Developed?” at www.iso.org/iso/en/stdsdevelopment/whowhenhow/how.html.
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 923
TABLE 3
Correlation Analysis between α and c Scores and Their Elements, as Well as Disclosure Measures
α Score
(0–4)
All Members
Corporate?
SIG?
Decisions Made
by Majority Rule?
Younger
Organization?
pppppCoefficient value Coefficient value Coefficient value Coefficient value Coefficient value
c Score (0–3)
Royalty-free
licensing?
Royalty-free or
RND
licensing?
Binding dispute
resolution?
Patent award
disclosure
required?
Patent
application
disclosure
required?
−0.53
−0.31
0.000
0.017
−0.36
0.005
−0.31
0.023
−0.18
0.190
0.01
0.987
−0.32
0.018
−0.43
0.001
−0.25
0.071
−0.27
0.050
Additional analyses. In this section, we look at three additional predictions of our model.
These relate to the relationships between disclosure requirements and licensing price, the
impact of limited competition between SSOs, and the consequences of the differing maturity of
standards.
First, as was discussed in Section 4 above, we also hypothesize a relationship between
disclosure and price. In particular, Proposition 2 suggests that within an equilibrium, a higher
licensing price was associated with more disclosure.
In Table 6, we look at the two most commonly encountered terms relating to licensing fees and
their relationship with the disclosure provisions. We find that the presence of a provision mandating
TABLE 4
Ordered Logit Regression Analysis of c Score, with α Score and Its Elements as Explanatory
Variables
Dependent Variable: c Score
α Score
−1.13
[0.30]∗∗∗
−1.36
[0.35]∗∗∗
SIG?
All members corporate?
Decisions made by majority rule?
Younger organization?
Dummies for technology included?
χ 2 -statistic
p-value
Log likelihood
Number of observations
N
17.41
0.000
−47.67
55
Y
23.89
0.000
−44.43
55
t-statistics are in brackets.
∗∗∗
denotes significant at the 1% confidence level; ∗ ∗ , 5%; ∗ , 10%.
C RAND 2007.
−1.85
[0.81]∗∗
−1.71
[0.67]∗∗
−0.74
[0.62]
−0.66
[0.60]
N
19.99
0.000
−46.38
55
−2.04
[0.83]∗∗
−1.79
[0.70]∗∗∗
−0.92
[0.64]
−1.09
[0.65]∗
Y
25.73
0.000
−43.51
55
C RAND 2007.
9.49
0.024
−15.20
48
6.51
0.011
−17.98
57
7.85
0.005
−30.56
57
9.88
0.020
−27.52
53
−0.97
[0.36]∗∗∗
Y
Royalty-Free or
RND Licensing?
−0.82
[0.32]∗∗∗
N
t Statistics are in brackets.
∗∗∗
denotes significant at the 1% confidence level; ∗ ∗ , 5%; ∗ , 10%.
−1.19
[0.56]∗ ∗
Y
−1.15
[0.53]∗∗
N
Royalty-Free
Licensing?
6.32
0.012
−13.59
55
−1.43
[0.72]∗ ∗
N
11.05
0.026
−11.22
55
−2.10
[1.03]∗∗
Y
Binding Dispute
Resolution?
Dependent Variable
1.77
0.183
−29.72
57
6.41
0.093
−27.40
57
−0.29
[0.30]
Y
Is Patent
Disclosure Required?
−0.39
[0.29]
N
Logit Regression Analysis of Elements of c Score and Disclosure Elements, with α Score as Explanatory Variable
Dummies for
technology included?
χ 2 Statistic
p Value
Log likelihood
Number of observations
α Score
TABLE 5
0.00
0.966
−30.07
55
0.01
[0.29]
N
3.25
0.355
−28.45
55
0.17
[0.32]
Y
Is Application
Disclosure Required?
924 / THE RAND JOURNAL OF ECONOMICS
CHIAO, LERNER, AND TIROLE / 925
TABLE 6
χ 2 Analyses of c Score Elements Regarding Price and Disclosure
Measures
Award Disclosure
Required?
Yes (%)
Royalty-free licensing?
RND licensing?
∗∗∗
5
77
No (%)
38∗∗∗
15∗∗
Application
Disclosure Required?
Yes (%)
0
85
No (%)
17
57∗
denotes significant at the 1% confidence level; ∗∗ , 5%; ∗ , 10%.
royalty-free licensing is negatively associated with the presence of a disclosure requirement,
whereas RND licensing is strongly associated with such a requirement. The pattern goes the
same way in both the analysis of the disclosure of patent awards and applications, but it is much
stronger in the former case.
The pattern of more disclosure being associated with higher licensing rates in Table 6 is
broadly consistent with the theoretical predictions above. The fact that the relationship is stronger
for patent awards is also consistent with our predictions. Lemma 1 states that if the disclosure
cost is high, then it is more likely that there will be pooling with no disclosure, making the
relationship overall weaker. Because it is plausible to assume that the disclosure cost is higher for
patent applications than for awards, this pattern is also expected.
Second, as noted above, an alternative hypothesis for the relationship between c and α is that
the patterns are a result of market power. It may be that some SSOs have few competitors. As a
result, they may be able to demand more concessions from sponsors while having a much more
user-orientated approach. As shown in the third subsection of Section 4, this argument, although
initially plausible, does not bear up under scrutiny. As Proposition 4 shows, when there are a
limited number of SSOs, it is by no means clear that the relationship between α and c will still
hold.
We address this issue in two ways. First, we rerun the regressions in Table 4, simply adding a
proxy for the market power of each SSO. The proxy we employ is the density of other SSOs in the
same technological subfield(s) as a given SSO. 32 If the sponsor can turn to many other SSOs, then
it is unlikely that the SSO can impose these types of requirements. We determine this measure
again through the ILI Infobase, using the classification scheme in the International Organization
for Standardization’s ISO Catalogue. 33 In general, the density of SSOs is quite high. The mean
SSO has 13.9 other SSOs in its subfield (with a median of 13.5).
We check to see whether once this control is added, the relationship between α and c still
holds. Then, we compare the goodness of fit in regressions when SSOs do and do not have
considerable competition.
Table 7 presents the results of the analysis. In the first four regressions, we examine the
impact of adding the measure of SSO density in the technological subfield. We find the measure
has little impact: across all four regressions (and numerous unreported ones), it is not statistically
or economically significant. As before, there is a strong negative association between c and α.
We then compare the goodness of fit in regressions using SSOs that did and did not have
market power. We divide the SSOs into those where there was above and below the median number
of other SSOs in their technological subfields. In the reported regressions (and in numerous
unreported ones), the goodness of fit is higher when we use SSOs located in subfields with many
other organizations: the absolute t statistics of the independent variables of interest are larger,
the χ 2 statistics are more statistically significant, and the log likelihoods are smaller. This is
32
For instance, if an organization is active in subfields A and B, and there are three and four active organizations
in A and B, respectively, then the market power index for this organization is (3 + 4)∗ 1/2.
33
Of course, the sponsor may have the option to create a new SSO, a possibility that our measure can only
imperfectly capture.
C RAND 2007.
926 / THE RAND JOURNAL OF ECONOMICS
TABLE 7
Ordered Logit Regression Analysis of c Score, with Market Power Proxy
Dependent Variable: c Score
Dividing Sample into Above
and Below Median SSO Density
Using Entire Sample
α Score
Above
−1.12
−1.35
[0.30]∗∗∗ [0.35]∗∗∗
SIG?
All members corporate?
Decisions made by
majority rule?
Younger organization?
−0.02
[0.06]
Dummies for technology
N
included?
χ 2 -statistic
17.53
p-value
0.000
Log likelihood
−47.61
Number of observations
55
Density of other SSOs
−0.01
[0.07]
Y
23.95
0.000
−44.41
55
Below
Above
Below
−2.15
[1.11]∗
−2.38
[1.13]∗∗
−0.37
[0.99]
−2.09
[1.07]∗
−35.79
[229.07]
−1.64
[1.12]
−1.45
[1.23]
−0.60
[0.94]
−1.55
−1.32
[0.56]∗∗∗ [0.54]∗∗
−1.84
[0.81]∗∗
−1.74
[0.68]∗∗∗
−0.69
[0.62]
−0.61
[0.61]
−0.04
[0.06]
N
20.36
0.001
−46.20
55
−2.08
[0.83]∗ ∗
−1.83
[0.71]∗∗∗
−0.82
[0.66]
−1.05
[0.65]
−0.04
[0.07]
Y
26.07
0.001
−43.35
55
Y
13.56
0.001
−18.91
29
Y
Y
Y
9.84
15.69
12.91
0.043
0.008
0.074
−24.04 −17.85 −22.51
26
29
26
t-statistics are in brackets.
∗∗∗
denotes significant at the 1% confidence level; ∗∗ , 5%; ∗ , 10%.
particularly striking in the last pair of regressions reported in Table 7. When using those SSOs
with above-median density, three of four c score elements are statistically significant; when using
those below the median density, none are. This pattern is consistent with the predictions of
Proposition 4.
The final analysis in Table 8 examines the relationship between the maturity of the standards
in the technology where the SSO is operating and the α measure. We believe that the best
indicator of opportunity is not the maturity of the project per se, but rather the collective state of
projects in that particular technological subfield. Over time, many of the substantial technological
uncertainties about standards in a subfield are likely to be resolved, increasing its attractiveness.
If a field is less developed, it would be more difficult for technical committees to come up with
standard proposals or drafts that take into account all technological implications, thus lengthening
TABLE 8
t-test and Ordered Logit Regression Analysis of α Score Explanatory Variable: Maturity Proxy
α if mature
α if not mature
t-statistic
p-value
Mature technology
χ 2 -statistic
p-value
Log likelihood
Number of observations
1.7
1.1
1.92
0.060
Dependent Variable: α Score
−1.00
[0.54]∗
3.45
0.063
−81.88
58
t-statistics are in brackets.
∗∗∗
denotes significant at the 1% confidence level; ∗ ∗ , 5%; ∗ , 10%.
C RAND 2007.
CHIAO, LERNER, AND TIROLE / 927
the approval time of standards in the field. Alternatively, there may be a higher option value to
sponsors from delaying their decision to participate, as suggested in Farhi, Lerner, and Tirole
(2005). We thus use the maturity level of standards in a technological field as a proxy for the
standard’s attractiveness to users.
We compute the maturity of the technological subfield(s) in which each standard operates
as follows. Following the procedure outlined above, we assign each organization to one or more
subfields as delineated in the ISO Catalogue. We then construct the maturity level for each
of the subfields by summing up the maturity ranks of all other standards in the subfield and
dividing the sum by the number of standards in each subfield. We use the ISO’s rating of maturity,
which indicates the maturity of that standard on a scale of 0 (preliminary stage) to 99 (standard
withdrawn after being implemented). We compute the average maturity of the standards in that
category. If there are no standards in a subfield, it receives the least mature rank. If an organization
spans across fields or subfields, we compute a simple average of the maturity levels for all relevant
fields or subfields. 34
Table 8 reports that there appears to be a positive correlation between maturity and α. The first
panel indicates that SSOs operating in subfields where standards are above the median maturity
tend to have a significantly higher α. The second panel presents an ordered logit regression,
with the α score of the SSO as the dependent variable. Once again, a higher maturity score is
significantly associated with a higher α, consistent with the theoretical predictions delineated
above.
7. Conclusion
Standard-setting organizations have received surprisingly little empirical scrutiny, despite
their economic importance and dynamism. This article seeks to address this omission, empirically
examining a cross-section sample of nearly sixty SSOs.
From Lerner and Tirole (2006), we expected a negative relationship between the extent to
which an SSO is oriented to technology sponsors and the concession level required of sponsors. In
this article, we indeed find a significant negative relationship, even when we control for industry
effects. We also expected the sponsor friendliness of the selected SSO to be positively associated
with the quality of a standard. The data reveal a statistically significant association between
sponsor friendliness and the maturity of a technological subfield in which the standard is located,
which we suggest should be a proxy for attractiveness.
Our theoretical analysis of disclosure requirements predicts that a higher licensing price
should be associated with more disclosure. Empirically, we find that the presence of a provision
mandating royalty-free licensing is negatively associated with the presence of a disclosure
requirement, whereas weaker “reasonable” licensing provisions are strongly associated with
such a requirement. Second, we show that in settings where there are only a limited number of
SSOs, the relationship between concessions and user friendliness may not hold. When we divide
our sample of SSOs into those where there were above and below the median number of other
SSOs in their technological subfield, we find that the relationship between user friendliness and
concessions is considerably tighter among SSOs located in classes with many other SSOs.
This work leaves a number of questions unexplored. One of the most intriguing of these has
to do with the dynamics of certification. Lerner and Tirole (2008) and the extensions discussed
here present a static model in which a one-time decision is made. In the real world, SSOs and
sponsors may employ more complex strategies: for instance, a sponsor may reapply to an SSO
after its initial application is rejected. (Farhi, Lerner, and Tirole, 2006 present a theoretical look
at these issues.) Understanding the dynamics of the certification process represents an important
empirical challenge.
34
In principle, we could have calculated a weighted average, but because we do not have information for the number
of standards of most of the organizations, we have not pursued this approach.
C RAND 2007.
928 / THE RAND JOURNAL OF ECONOMICS
Appendix
Proofs of Proposition 2–4 follow.
Proof of Proposition 2. Compare the two programs
π D (d, a) =
max
{b∗ | b∗ ≥m −1 (−(a+H ))}
{[1 − F(b∗ )] [π0 + (a + H ) + m(b∗ ) − d]},
and
π N D (0, a) =
max
|
{b̂ b̂ ≥m −1 (−a)}
{[1 − F(b̂)] [π0 + a + m(b̂)]}.
(i) Let us first demonstrate that disclosure is associated with a higher price in a separating equilibrium. Let us assume
a contrario that PND > PD . This inequality can arise only if PND > 0. Because PND > 0, then α ND = 1 and so the
cutoff b̂ is the efficient cutoff. By contrast, α D ≤ 1 (with equality if PD > 0) and so the cutoff b∗ is either efficient or
socially too high. Thus we have b∗ ≥ b̂. And so
P D − P N D = [a + H + mb∗ ] − [a + m(b̂) > 0]
(ii) Let us show that α D ≥ α ND . This is clearly the case when P D > 0, as then α D = 1 and α ND ≤ 1. So, assume that
P D = PND = 0. Using the users’ and the SSO’s indifference equations, one then gets
a + H + m(b∗ ) = a + m(b̂) = 0
and
(a + H ) + b∗ + α D [π0 − d] = a + b̂ + α N D π0=0 ,
implying that
(α D − α N D )π0 = [m(b∗ ) − b∗ ] − [m(b̂) − b̂] + α D d,
which together with m < 1 and b∗ < b̂ yields α D > α ND .
Proof of Proposition 3. Recall that d ∗ is given by
π D (d ∗ , a) = π N D (ρ, a).
Let P D and PND denote the prices under disclosure, and nondisclosure and pooling, respectively (beware that PND is
not the same as in Proposition 2, as we are now looking at pooling rather than separating). The same proof as for
Proposition 2 shows that P D ≥ PND .
Suppose, first, that P D = PND = 0. Then, for d = d ∗ ,
[1 − F(m −1 (−a − H ))](π0 − d ∗ ) = [1 − F(m −1 (−a − ρ H ))]π0
and so
∂(π N D − π D )
=
∂a
1 − F(b̂)
= [1 − F(b̂)]π0 [
(using the fact that m =
f
1−F
m(b̂)−b̂
π0 −
1m(b̂) − b̂
−
1
1 − F(b∗ )
m(b∗ )−b∗
(π0 − d ∗ )
m(b∗ ) − b∗ ]
(m − b)). Because [ m(b) − b ] is decreasing and b̂ > b∗ ,
∂(π N D − π D )
> 0.
∂a
Thus, there is disclosure for a smaller set of disclosure costs as a increases.
Suppose, next, that P D ≥ PND > 0. Then π D = π ND requires that d ∗ = (1 − ρ)H and b̂ = b∗ . Furthermore,
∂(π N D − π D )
= [1 − F(b̂)] − [1 − F(b∗ )] = 0.
∂a
Finally, assume that P D >PND = 0. Then
∂(π N D − π D )
π0 − d ∗ + P D
−1 .
∝ [1 − F(b̂)]
∂a
m(b̂) − b̂
Using the efficiency condition, however,
π0 − d ∗ + P D = m(b∗ ) − b∗ .
Finally, m < 1 and b∗ < b̂ implies that
C RAND 2007.
∂(π
ND
− π D)
∂a
> 0.
CHIAO, LERNER, AND TIROLE / 929
Proof of Proposition 4. Consider a sponsor of a technology with known attractiveness a. Let α ∗ denotes his “ideal SSO,”
and (b∗ , c∗ ) denote the corresponding cutoff and concession level:
⎧
a + m(b∗ ) + c∗ = 0
⎪
⎪
⎪
⎪
⎨
a + b∗ + c∗ + α ∗ π (c∗ ) = 0
⎪
⎪
⎪
⎪
⎩
1 + α ∗ π (c∗ ) = 0.
An SSO with type α = α ∗ attempts to offer as high a surplus [1 − F(b)]π (c) as it can. If α > α ∗ , the SSO lacks credibility
in the eyes of users and must make a high concession to gain sufficient credibility:
c > c∗ .
To see this, suppose, to the contrary, that c ≤ c∗ . Then letting b denote the cutoff for (c, α),
a + b + c + απ (c) = 0
and
a + m(b) + c ≥ 0
if the standard is to be adopted by users. Therefore
m(b) − b ≥ απ (c) > α ∗ π (c∗ ) = m(b∗ ) − b∗ ,
which, together with m < 1, yields b < b∗ . However, then
a + m(b) + c < a + m(b∗ ) + c∗ = 0,
and so the standard is not adopted by users after all. Note also that for α > α ∗ , the two constraints in program (I) must be
binding; otherwise, the first-order condition would yield
π
f
+
(1 + απ ) = 0.
π
1− F
∗
∗
∗
(A1)
∗
However, α > α , c >c , π < 0, and 1 + α π (c ) = 0 imply that the left-hand side of (A1) is strictly negative.
Similarly for α < α ∗ , but “not too small,” the two constraints in program (I) must be binding. For α close to α ∗ ,
(A1) would yield c far below c∗ , which by continuity cannot be the solution. 35
Now, when the two constraints are binding,
a + m −1 (−a − c) + c + απ (c) = 0,
and so
dc
=
dα
1
m
π
.
− (1 + απ )
Now 1 + α π < 0 for α > α ∗ , and (1 + α π ) small for α smaller than, but in neighborhood of, α ∗ . Hence
dc
dα
> 0.
References
BESEN, S. AND JOHNSON, L. Compatibility Standards, Competition and Innovation in the Broadcast Industry. Santa
Monica, CA: The RAND Corporation, 1986.
——— AND SALONER, G. “The Economics of Telecommunications Standards.” In R.W. Crandall and K. Flamm, eds.,
Changing the Rules: Technological Change, International Competition, and Regulation in Telecommunications,
1989.
BULOW, J. AND KLEMPERER, P. “The Generalized War of Attrition.” American Economic Review, Vol. 89 (1999), pp.
175–189.
CARGILL, C. “Uncommon Commonality: A Quest for Unity in Standardization.” In Sherrie Bolin, ed., The Standards
Edge, Ann Arbor, MI: Bolin Communications, 2002, pp. 29–39.
——— AND BOLIN, S. “Standardization: A Failing Paradigm.” In Shane Greenstein and Victor Stango, eds., Standards
and Public Policy, New York: Cambridge University Press, 2007, forthcoming.
35
To show that the highest gross sponsor payoff cannot decrease discontinuously as α decreases, suppose that SSO
α offers concession c so that the standard is adopted if recommended by the SSO (if not, then obviously the payoff cannot
decrease as it is equal to 0): a + m(b) + c ≥ 0, a + b + απ (c) = 0. Now consider SSO (α − dα) with dα > 0. It is easy
to see that for concession c + dc, the cutoff is b + db with db = π (c)dα − απ (c) dc, and that the users still adopt as
long as dc ≥ −{[π (c)m (b)dα]/[1 − απ (c)m (c)]}.
C RAND 2007.
930 / THE RAND JOURNAL OF ECONOMICS
COWIE, M.G. AND LAVELLE, J.P. “Patents Covering Industry Standards: The Risks to Enforceability Due to Conduct before
Standard-Setting Organizations.” AIPLA Quarterly Journal, Vol. 20 (2002), 95ff.
DOLMANS, M. “Standards for Standards.” Testimony to the U.S. Department of Justice Antitrust Division/Federal Trade
Commission Hearings on Competition and Intellectual Property Law and Policy in the Knowledge-Based Economy,
Washington, DC, 2002.
ERTURK, E., LANSFORD, B., AND MUSCARELLA, C. “Patent Announcements and Corporate Value.” Unpublished Working
Paper, Pennsylvania State University and Northwestern University, 2006.
FARHI, E., LERNER, J., AND TIROLE, J. “Certifying New Technologies.” Journal of the European Economic Association,
Papers and Proceedings, Vol. 3 (2005), pp. 734–744.
———, ———, AND ———. “Certification, Second Chances, and Disclosure.” Unpublished Working Paper.
FARRELL, J. “Choosing the Rules for Formal Standardization.” Unpublished Working Paper, University of California–
Berkeley, 1996.
——— AND SALONER, G. “Standardization, Compatibility, and Innovation.” RND Journal of Economics, Vol. 16 (1985),
pp. 70–83.
——— AND ———. “Coordination through Committees and Markets.” RND Journal of Economics, Vol. 19 (1988), pp.
235–252.
——— AND SHAPIRO, C. “Standard Setting in High-Definition Television.” Brookings Papers on Economic Activity:
Microeconomics (1992), pp. 1–77.
FELDMAN, R.P., REES, M.L., AND TOWNSHEND, B. “The Effect of Industry Standard Setting on Patent Licensing and
Enforcement.” IEEE Communications, Vol. 38 (2000), pp. 112–116.
GANDAL, N., GANTMAN, N., AND GENESOVE, D. “Intellectual Property and Standardization Committee Participation in
the U.S. Modem Industry.” In Shane Greenstein and Victor Stango, eds., Standards and Public Policy, New York:
Cambridge University Press, 2007, forthcoming.
HEMENWAY, D. Industrywide Voluntary Product Standards. Cambridge, Mass.: Ballinger Publishing, 1975.
JAFFE, A.B. AND LERNER, J. Innovation and Its Discontents. Princeton, NJ: Princeton University Press, 2004.
KIPNIS, J. “Beating the System: Abuses of the Standards Adoption Process.” IEEE Communications, Vol. 38 (2000), pp.
102–105.
KRATHWOHL, D.R. Methods of Educational and Social Science Research: An Integrated Approach, 2d ed. New York:
Longman, 1997.
LEHR, W.H. “Compatibility Standards and Industry Competition: Two Case Studies.” Economics of Innovation and New
Technology, Vol. 4 (1996), pp. 97–112.
LEMLEY, M.A. “Intellectual Property Rights and Standard Setting Organizations.” California Law Review, Vol. 90 (2002),
pp. 1889–1990.
LERNER, J. AND TIROLE, J. “A Model of Forum Shopping.” American Economic Review (2006).
LICHTMAN, D. “Defensive Suspension in Standard-Setting Organizations.” Unpublished Working Paper, University of
Chicago, 2005.
MUELLER, J.M. “Patent Misuse through the Capture of Industry Standards.” Berkeley Technology Law Journal, Vol. 17
(2002), pp. 623–684.
ROSENKOPF, L., METIU, A., AND GEORGE, V.P. “From the Bottom Up? Technical Committee Activity and Alliance
Formation.” Administrative Science Quarterly, Vol. 46 (2001), pp. 748–772.
RYSMAN, M. AND SIMCOE, T. “Patents and the Performance of Voluntary Standard Setting Organizations.” NET Institute
Working Paper no. 05-22, 2005.
SIMCOE, T. “Delay and de Jure Standards: Exploring the Slowdown in Internet Standards Development.” In Shane
Greenstein and Victor Stango, eds., Standards and Public Policy, New York: Cambridge University Press, 2007,
forthcoming.
SIRBU, M. AND ZWIMPFER, L. “Computer Communications Standards: The Case of X.25.” IEEE Communications, Vol.
23 (1985), pp. 35–45.
UPDEGROVE, A. “Intellectual Property Rights Policies.” Unpublished Working Paper, 2006. Available from
www.consortiuminfo.org/ipr.
U.S. FEDERAL TRADE COMMISSION. Standards and Certification. FT 1.2:St 1 and FT 1.2:St 2/2/Final. Washington, DC:
Government Printing Office, 1978, 1983.
WEISS, M.B.H. AND SIRBU, M. “Technological Choice in Voluntary Standards Committees: An Empirical Analysis.”
Economics of Innovation and New Technology, Vol. 1 (1990), pp. 111–133.
C RAND 2007.